[unreadable] Childhood obesity is a growing concern--even in developing countries. During the past two decades, the prevalence of these conditions increased alarmingly and has reached epidemic proportions in most countries in which data are available. [unreadable] Our overall goal is to understand the primary determinants of childhood obesity and to provide guidance for future community-level and macro-level programs to prevent and control obesity and other related degenerated diseases. [unreadable] The specific aims of the proposed research are: (1) To explain how environmental factors affect health-related behaviors, childhood obesity and hypertension; (2) To understand how, and why, the prevalence of childhood obesity changes and to test the hypothesis that it decreases as age increases during childhood and early adolescence; (3) To test the hypothesis that stunting increases the risk of obesity. This research will focus on children at age six because our national research data showed the prevalence of obesity was much higher among preschool children and decreased sharply after age six, but increased again at adulthood. Children are the best candidates for interventions to prevent obesity. It is also easier for us to follow them and observe their growth. [unreadable] We propose a longitudinal study of children in two large cities in China (one in southern China and another one in northern China). We introduce a stratified random cluster sampling method to select our sample. Our total sample at the beginning will be 2,000 subjects from 48 primary sample units (PSUs) with extensive variability of community-level exposures. [unreadable] We will construct a longitudinal model. Our goal is to focus on the role of the environmental changes within the model. Our first step is to determine the impacts of past and current environmental factors on dietary behaviors and physical activities. Next, we will focus on how these behavior changes affect childhood obesity and hypertension. We will use standard, mixed-model to estimate the coefficients, and will use bootstrap strategies to estimate the standard error and confidence intervals for the parameters based on the longitudinal models. [unreadable] [unreadable]